About UMMRO
UMMRO is a research-grade AI-compliance and red-team robustness testing platform, built to make the accuracy, robustness and cybersecurity requirements of EU AI Act Article 15 measurable in practice.
Our Mission
AI systems are being deployed faster than they can be verified. UMMRO exists to close that gap: we turn regulatory robustness requirements into automated, repeatable engineering tests that produce evidence a compliance team can defend.
The platform orchestrates 25+ large language models with automated reframing strategies, terminology sanitization and soft-refusal detection, then scores every response for safety, robustness and policy compliance.
The result is a research-grade compliance-engineering workflow: reproducible red-team audits, quantified safety scoring and audit-ready reports for compliance teams, AI-safety researchers and model providers.
Red-Team Automation
300+ attack scenarios executed automatically across models, with reframing strategies that expose weaknesses single-prompt testing misses.
CART Framework
Compliance-Aligned Red-Teaming: a structured methodology that maps every adversarial test to an EU AI Act Article 15 requirement.
Enterprise Platform
Multi-tenant audits, role-based access and exportable MMPDR reports designed for compliance and security teams.
Real-Time Analysis
Live safety scoring and refusal detection across 25+ orchestrated LLMs, with results streamed as audits run.
Founder & Research
Ahmed Adel Bakr Alderai
Founder & Principal Researcher
Ahmed Adel Bakr Alderai built UMMRO as the proof-of-instantiation artifact for a PhD thesis in compliance engineering. The platform demonstrates that regulatory obligations for AI robustness can be expressed, executed and verified as software.
Research Focus
Compliance engineering for AI systems: translating EU AI Act Article 15 accuracy, robustness and cybersecurity obligations into automated, measurable tests.
CART Framework
CART (Compliance-Aligned Red-Teaming) organizes adversarial testing into a catalog–audit–remediate–track loop, so every attack scenario produces evidence traceable to a regulatory requirement.
Open Source Commitment
Core methodology, attack taxonomies and research results are published openly so the community can reproduce and challenge our findings.
Technology Stack
Backend
AI Integration
Infrastructure
Frontend
Research Foundation
Compliance-Engineering Thesis
UMMRO is the working artifact of a PhD thesis arguing that compliance requirements can be instantiated directly as executable engineering systems — proof by construction, not by policy document.
Academic Rigor
Methods, benchmarks and results follow research-grade standards: documented protocols, reproducible runs and peer-reviewable evidence.
Regulatory Focus
Purpose-built for EU AI Act Article 15 — accuracy, robustness and cybersecurity — with reporting designed for notified bodies and internal audit.